3 ms·
I don't know if it's the hardest, but very recently I had a problem that drove me nuts: I was doing some consulting for a company that designs chips, and they a
by lisper 8y ago
I don't know if it's the hardest, but very recently I had a problem that drove me nuts: I was doing some consulting for a company that designs chips, and they assigned me a project to take a nearly completed design and augment it with additional wires to route extra power to some parts of the chip that were running on the hairy edge of not having enough current. So I had to read in the design spec, search it for available space, and generate new wire geometries for the power grid. This sounds straightforward, but there were two things that made is incredibly hard. First, the design that I was reading in was the output of a routing tool that was provided by a third-party vendor, and it was producing some really crazy shit. Not even the senior design engineers could understand some of the things that the router was doing. And second, there are hundreds of design rules that the new geometries had to conform to. Most of them are not relevant, but about a dozen or so are, and some of them involve incredibly complicated interactions of multiple geometries, occasionally across multiple metal layers. Trying to keep track of all that and make it all run in a reasonable amount of time was quite challenging, to say the least. But the worst part was that the development cycle involved a manual step where the output of my code had to fed into yet another third-party tool to see if I had violated any of the design rules, and the output of that tool was graphical. And I was not able to use the tool because of licensing restrictions! So I had to take my output, give it to someone else, have them run the tool, look at the results, and send me a report of how many rule violations there were, where they were, and what rules were being violated. That took several hours, sometimes multiple days.